From: A Grassmann graph embedding framework for gait analysis
Parameter | Description | Values/range of values that yield good performance in general |
---|---|---|
k | Number of neighbors for the local-based graph construction approach | 3 |
t | Constant in the heat kernel function | {1, 2, …, 10} |
q | The number of eigenvectors selected for the projection matrix | About 30% of the smallest eigenvectors for Equation 11; about 30% of the smallest eigenvectors for Equation 17 |
σ | A regularization term for solving the generalized eigenvalue problems (11) and (17) which can improve the result | {1, 2, …, 10} |
λ | A regularization parameter for the l1_ls procedure for SR | 0.1 |